SDS DASCA Senior Data Scientist Free Practice Exam Questions (2025 Updated)

Social Network Analysis (SNA) is a powerful analytical method that applies graph theory to study relationships among entities (people, organizations, computers, etc.).

Option A: Correct. SNA is indeed an example of graph analysis because it models entities as nodes and their relationships as edges/ties.

Option B: FALSE. SNA is not an example of trend analysis. Trend analysis focuses on temporal patterns (time series), while SNA is structural and relational.

Option C: Correct. SNA investigates structures such as communities, influencers, and information diffusion in networks.

Option D: Correct. The characterization of nodes and edges is central to SNA.

Option E: Incorrect, since we’ve identified Option B as false.

Thus, the false statement is Option B.

Data wrangling (also called data munging) refers to transforming raw, messy, or unstructured data into a clean and structured format suitable for analysis.

Option A: Correct. Raw data often contains missing values, duplicates, or irregular formats. Wrangling prepares it for conventional analytics and machine learning.

Option B: Incorrect. Wrangling does not involve "modified meaning"; it focuses on cleaning, structuring, and integrating.

Option C: Incorrect, since only A is correct.

Option D: Incorrect, because wrangling is explicitly described in A.

Thus, the correct answer is Option A.

The correct answer isPyMC(Option B).

PyMC is an open-source Python library widely used forBayesian statistical modelingandprobabilistic machine learning. It provides a robust framework for defining and fitting Bayesian networks to real data usingMarkov Chain Monte Carlo (MCMC)sampling techniques, as well asvariational inferencemethods. This makes it a powerful tool for data scientists who want to work withuncertainty modeling,probabilistic inference, andcausal reasoningin complex datasets.

Let’s clarify the other options to avoid confusion:

Option A: SciLib– There is no standard Python library by this name that is related to Bayesian networks. (It may be confused withSciPyorSciKit-Learn, but those are not specialized for Bayesian inference.)

Option C: MyLib– This is not a recognized Python package in the data science ecosystem.

Option D: MCMC– While Markov Chain Monte Carlo is thetechniqueused in Bayesian estimation, it is not a standalone library. Instead, PyMC implements MCMC as part of its computational framework.

Option E: SCIMC– No such Python library exists; it appears to be a distractor.

PyMC’s primary strength is its ability to let data scientists define models in aprobabilistic programming style, making it easier to represent uncertainties and hidden variables in data. This aligns with DASCA’s emphasis on ensuring data scientists understand bothstatistical foundationsand thetools required to implement them programmatically.

In practice, PyMC is often used in applications such as:

Forecasting(e.g., time series with uncertainty bounds)

Causal inference(estimating hidden relationships in data)

Risk modeling(finance, healthcare, or supply chain domains)

Machine learning with uncertainty quantification

Thus,PyMCis the correct library for fitting Bayesian networks in Python.

Natural Language Processing (NLP) is the field of machine learning that deals with human language understanding and generation. Applications include:

Option A (Spam detection): NLP techniques classify emails based on text patterns and context.

Option B (Sentiment analysis on Twitter): NLP models analyze textual data to extract emotions, opinions, and trends.

Option C (Topic modeling): NLP clustering and probabilistic models (e.g., LDA) classify documents by semantic similarity.

Since all are valid NLP applications, the correct answer is Option E (All of the above).

A Statement of Work (SOW) is a formal document that defines the scope, objectives, deliverables, timeline, and expectations of a project. In data science and IT projects, it ensures:

Clarity of scope (Option A): Everyone understands exactly what work should be done.

Clear priorities (Option B): It defines what is most critical for success.

Realistic expectations (Option C): It aligns stakeholders by setting measurable and achievable goals.

Since all of these are essential purposes of an SOW, the correct answer is Option D (All of the above).

Python supports multiple libraries for scientific computing and data analysis, but the primary machine learning library is:

Scikit-learn (Option B): Provides a wide range of machine learning algorithms for classification, regression, clustering, model evaluation, and preprocessing. It is the core ML library in Python.

NumPy (Option A): Provides numerical computing and array operations, essential for ML but not a machine learning library itself.

Matplotlib (Option C): Used for data visualization.

SciPy (Option D): Supports scientific computing and numerical methods, not focused on ML models.

Therefore, the correct answer is Option B (Scikit-learn).

The Big Data Vision Workshop is an early-phase framework designed to help organizations shape their data-driven transformation journey. It is particularly beneficial when:

Option A: Organizations want to leverage big data but lack clarity on where to start.

Option B: Organizations already have large volumes of data but struggle to derive monetization strategies from it.

Option C: Organizations want to identify use cases where data and analytics can enhance or even redefine their business models.

Since all three statements apply, the correct answer is Option E (All of the above).

Unsupervised learning is a machine learning approach where no labeled outputs are provided. The algorithm discovers patterns or structures directly from raw data.

Option A (Detecting regularities): Correct. Unsupervised learning identifies hidden structures such as clusters, associations, and dimensionality reductions (e.g., k-means clustering, PCA).

Option B (Detecting irregularities): Correct. Outlier detection is also a part of unsupervised learning, often used in anomaly detection (e.g., fraud detection, intrusion detection).

Option C: Correct, since unsupervised learning helps detect both regularities (clusters, groups) and irregularities (outliers, anomalies).

Thus, the correct answer is Option C (Both A and B).

Maximum Likelihood Estimation (MLE) is a statistical method used to estimate the parameters of a model by maximizing the likelihood function — i.e., finding the parameters that make the observed data most probable.

Option A: Correct. MLE provides a framework for a large class of problems in data science, including regression, classification, generative models, and probabilistic inference.

Option B: Incorrect — it applies to many problems, not just a small subset.

Option C & D: Incorrect. HDFS (Hadoop Distributed File System) is a storage technology, unrelated to MLE.

Option E: Incorrect because C is invalid.

Thus, the correct answer is Option A (Large class of problems in Data Science).

Microservices architecture is a software design approach where applications are developed as a collection of small, loosely coupled, and independently deployable services.

Option A: Correct. Each microservice runs independently and has its own lifecycle, allowing scalability and modularity.

Option B: Correct. Each service can be treated as an independent project with its own repository, build pipeline, and deployment.

Option C: Correct. Microservices can be developed in different programming languages or frameworks, chosen based on the best fit for each specific requirement.

Therefore, the correct answer is Option D (All of the above).

Scrum is a framework used to implement Agile principles. Agile itself is the overarching philosophy or mindset, while Scrum is one of the most popular frameworks that apply Agile values in practice.

Option A (Correct): Scrum is indeed a subset of Agile. Agile defines the principles (from the Agile Manifesto), and Scrum provides the structure (roles, artifacts, ceremonies).

Option B: Incorrect. Agile is broader and not a subset of Scrum.

Option C: Incorrect. Scrum and Agile are not the same; Agile is the philosophy, Scrum is a methodology under Agile.

Option D: Incorrect because Option A is valid.

Thus, the correct answer is Option A: Scrum is a subset of Agile.

Cluster management tools help in orchestrating and monitoring large-scale distributed computing environments.

Zettaset Orchestrator (A): Commercial tool for Hadoop cluster management.

Apache Mesos (B): A cluster manager that abstracts CPU, memory, and storage to enable fault-tolerant distributed systems.

Apache Ambari (C): An open-source tool for provisioning, managing, and monitoring Hadoop clusters.

Apache Hadoop (D): Not a cluster management tool. Hadoop is a framework for distributed storage and processing (HDFS + MapReduce), not a management tool.

Thus, the correct answer is Option D (Apache Hadoop).

Machine learning (ML) can be broadly divided into three main paradigms:

Supervised Learning (Option A):

Data includes labeled outputs (e.g., classification, regression).

Goal: Learn a mapping from input to output.

Unsupervised Learning (Option B):

Data has no labels.

Goal: Discover hidden patterns (e.g., clustering, dimensionality reduction).

Reinforcement Learning (Option C):

Agent interacts with an environment and learns by maximizing cumulative rewards through trial and error.

Used in robotics, game AI, and autonomous systems.

Since all three categories are valid, the correct answer is Option D (All of the above).

Customer Lifetime Value (CLTV) is a predictive metric estimating the total revenue a business can reasonably expect from a customer during their entire relationship.

Statement i: Correct. Many organizations calculate CLTV using historic transactional data, often looking at sales records over the past 12–18 months to establish baselines.

Statement ii: Correct. The primary purpose of CLTV is to help marketing, sales, and retail teams understand customer value, enabling them to allocate budgets effectively for retention, promotions, and personalized marketing.

Thus, both statements are correct → Option C (Both i and ii).

A Markov chain is a stochastic process describing a sequence of possible events, where the probability of each event depends only on the state attained in the previous step (the Markov property).

Option A: Incorrect. The variables are random, not fixed.

Option B: Correct. Markov chains represent the state of a system at sequential time points.

Option C: Correct. A Markov chain is indeed a sequence of random variables {X1, X2, …} that satisfy the Markov property.

Option D: Incorrect, since A is wrong.

Option E: Correct, because both B and C are valid.

Thus, the correct answer is Option E (Both B and C).

"By" analysis is one of the foundational approaches recommended in the DASCA Data Scientist Knowledge Framework for structuring problem-solving in data science. The purpose of "By" analysis is to enable data scientists and business stakeholders to think beyond obvious data correlations and uncover deeper drivers of business outcomes.

At its core, the technique reinforces the discipline ofthinking like a data scientist(Option A). This involves reframing business questions into analytical structures and asking “What drives this metricbywhich factors?” For example, customer churn might be analyzedbydemographics, purchase behavior, or service usage. This structured mindset is critical for ensuring scientific rigor in business problem analysis.

In addition, "By" analysis emphasizes collaboration betweenSubject Matter Experts (SMEs)andData Science teams(Option B). SMEs bring contextual domain knowledge, while data scientists bring analytical and statistical expertise. Together, they brainstorm possible explanatory variables or metrics that could become strong predictors of business performance.

Furthermore, the process provides acollaborative bridgebetween business and technical stakeholders (Option C). It ensures that the exploration of data is not isolated in silos but is grounded in both domain insights and advanced analytical methods. This alignment is crucial for building models that are not only technically sound but also relevant and actionable in real-world business contexts.

Since Options A, B, and C are correct and complementary, the best choice isOption E: All of the above.

A Tensor is a fundamental data structure in modern machine learning frameworks (e.g., TensorFlow, PyTorch). It is best described as a generalization of vectors and matrices to potentially higher dimensions.

Option A: Correct. Tensors typically store numeric values (commonly floating-point numbers) in structured formats.

Option B: Correct. A tensor can have any number of dimensions (rank). For example:

A scalar is a 0-D tensor.

A vector is a 1-D tensor.

A matrix is a 2-D tensor.

Higher-rank tensors can represent images, videos, or multidimensional datasets.

Option C: Correct. Tensors are explicitly designed to allow linear algebra operations, which are the foundation of deep learning computations (matrix multiplications, dot products, etc.).

Therefore, since all three statements are true, the correct answer is Option E (All of the above).

Optical Character Recognition (OCR) is the process of automatically recognizing and converting different types of documents — such as scanned paper documents, PDFs, or images — into editable and searchable text.

OCR systems use Machine Learning (ML) and Computer Vision techniques to detect and classify patterns of characters in images.

Algorithms like Convolutional Neural Networks (CNNs) are commonly used for image-based OCR.

While OCR may indirectly contribute to data mining or big data workflows, the core application is based on machine learning, where models are trained to classify and recognize text patterns.

Thus, OCR is primarily a Machine Learning application, making Option B correct.

ARIMA stands for AutoRegressive Integrated Moving Average, one of the most widely used models for time series forecasting.

AutoRegressive (AR): Model uses past values of the variable to predict future values.

Integrated (I): Differencing is applied to make the time series stationary.

Moving Average (MA): Model incorporates past forecast errors into predictions.

Option B: Correct — autoregressive + moving average is part of ARIMA’s name.

Options A, C, D: Incorrect because these terms are not recognized statistical modeling frameworks.

Option E: Incorrect, since only B is valid.

Thus, the correct answer is Option B.